基于机器学习的含缺陷PE管道承载能力研究

Study on the Carrying Capacity of the PE Pipeline with Defects Based on Machine Learning

通过拉伸试验与DIC方法拟合得到初始弹性模量和屈服应力等相关参数,采用有限元方法,获取管道在内压、弯矩、轴向力联合载荷作用下的承载能力。在此基础上构建了包含相对深度(C/T),相对轴向长度(2A/√RT),相对周向角度(2θ/π)和无量纲参数(c)等4个参数的含局部减薄缺陷PE管道的BP神经网络模型,并结合GA优化BP神经网络模型进行对比分析。可以发现,模型的预测值与模拟的结果较为一致,表明采取GA优化BP神经网络模型的方法是可行的,为含局部减薄缺陷PE管道的智能化安全评价提供了有效方法。

PE pipeline is widely used in urban gas pipe network.The carrying capacity of PE pipe with local thinning defects is an important means to ensure its safe operation.Firstly,the initial elastic modulus and yield stress are fitted by the finite element method to obtain the carrying capacity of internal pressure,bending moment and axial force,the BP neural network model including relative depth C/T,relative axial length 2A/√RT,relative circumferential angle 2θ/π and dimensionless parameters c,and combined with GA optimized BP neural network model.It can be found:the predicted value of the model is relatively consistent with the simulation results,which shows that the method of GA optimization of the BP neural network model is feasible,which provides an effective method for the intelligent safety evaluation of the PE pipeline with local thinning defects.